Question

我有多行有开始时间＆amp;结束时间彼此重叠的列。

我需要使用SQL找到不同的时间间隔。

示例数据：

(6 -> 7)
(6.30 -> 6.45)
(8 -> 9)
(8.30 -> 9.30)

输出：

(6 -> 7)
(8 -> 9.30)

Answer 1

Vertica拥有非常强大的时间序列＆＃34;和＆＃34;条件事件＆＃34;分析功能。你的问题可以通过这种方式轻松解决......

假设这是你的起点：

SQL> select * from otest ;
        t1          |         t2          
--------------------+--------------------
2016-03-04 06:00:00 | 2016-03-04 07:00:00
2016-03-04 06:30:00 | 2016-03-04 06:45:00
2016-03-04 08:00:00 | 2016-03-04 09:00:00
2016-03-04 08:30:00 | 2016-03-04 09:30:00
(4 rows)

t1是开始时间戳，t2是结束时间戳。您所要做的就是：

SQL> select 
         min(a.t1), 
         max(a.t2) 
     from ( 
         select 
            t1, 
            t2, 
            conditional_true_event ( t1 >= lag(t2) ) 
               over ( order by t1 ) as cte 
         from otest ) a  
     group by cte 
     order by 1 ;

        min         |         max         
--------------------+--------------------
2016-03-04 06:00:00 | 2016-03-04 07:00:00
2016-03-04 08:00:00 | 2016-03-04 09:30:00
(2 rows)

Answer 2

我会评论Mauro's，但我没有代表。不幸的是，他的回答并没有考虑当你有两个以上的重叠期时会发生什么。

这是我的解决方案：

--create the table for the purposes of this demo
drop schema if exists TEST1 cascade;     
create schema if not exists TEST1;
drop table if exists      TEST1.otest;
create table if not exists TEST1.otest(t1 datetime, t2 datetime);

--create some example data
--example where 2nd period is entirely inside the first
insert into TEST1.otest(t1, t2) select '2016-03-04 06:00:00' ,'2016-03-04 07:00:00';
insert into TEST1.otest(t1, t2) select '2016-03-04 06:30:00' ,'2016-03-04 06:45:00';

--example of multiple consecutive periods
insert into TEST1.otest(t1, t2) select '2016-03-04 08:00:00' ,'2016-03-04 09:00:00';
insert into TEST1.otest(t1, t2) select '2016-03-04 08:15:00' ,'2016-03-04 08:25:00';
insert into TEST1.otest(t1, t2) select '2016-03-04 08:26:00' ,'2016-03-04 08:27:00';
insert into TEST1.otest(t1, t2) select '2016-03-04 08:28:00' ,'2016-03-04 08:29:00';
insert into TEST1.otest(t1, t2) select '2016-03-04 08:30:00' ,'2016-03-04 09:30:00';

--example of another overlapping period extending the end time
insert into TEST1.otest(t1, t2) select '2016-03-04 10:00:00' ,'2016-03-04 10:30:00';
insert into TEST1.otest(t1, t2) select '2016-03-04 10:15:00' ,'2016-03-04 10:45:00';

--query syntax
with i as (select * from TEST1.otest) 
,i2 as (select * ,max(t2) over (order by t1) as maxT2 from i)
,i3 as (select *, lag(i2.maxT2) over (order by t1) as laggedMaxT2 from i2)
,i4 as (select *, conditional_true_event(i3.t1 > i3.laggedMaxT2) over (order by t1) as grouper from i3)
select min(t1) as collapsedT1, max(t2) as collapsedT2 from i4 group by grouper
order by collapsedT1;

--results
    collapsedT1         |collapsedT2         |
--------------------|--------------------|
2016-03-04 06:00:00 |2016-03-04 07:00:00 |
2016-03-04 08:00:00 |2016-03-04 09:30:00 |
2016-03-04 10:00:00 |2016-03-04 10:45:00 |

编辑：如果您的数据按其他列分类，请记住将分区子句添加到max，conditional_true_event和lag分析中，否则您可能会获得非确定性结果。

重叠时间戳的SQL查询

2 个答案: